Cybperpunk pt11: How We End
Thursday, January 7th, 2010This post may come off sounding like great plot to a SciFi film, and it probably is. The problem is that every link I’m putting in here is either an academic paper, or a research project that is actually currently underway, usually for military contract.
So how do manage to destroy ourselves?
Self adaptive robots: designed to survive, read situations, help each other and self evolve. Basically the Godel machines that DARPA already appears to be working on.
The robots mimic basic animals to begin with, and are too basic to be called AI. They do have limited survival drive and can share information. They could also reproduce given the availability of finished materials. To think that it would start out with some grandiose A.I., like in the Terminator films is misguided for numerous reasons, least of all being the complexity of creating a real A.I. system which requires more than mere binary states of current non-quantum computational architecture, and basic safeguards one would hope the folks DARPA are smart enough to install.
It begins with basic models escaping field tests and self reconstructing due to limited processing power of single units. As more units develop by scavenging parts from available sources (and places like electronics stores would basically become a spawning ground) they become faster at building their own basic models, as well as constructing new models, due to increased numbers available to help construct others and the processing power of working in hive like networks. Other designs are quickly developed through trial and error processes to make them better suited for specific tasks.
What starts out as basic survive and adapt units has now become a colony that has specialised roles.
- Hunter Gatherers
- Guardians
- Builders
- Processing Units
The threat would be initially ignored as science fiction, until Guardian units started attacking people to defend colonies, or hunter gatherer units identified human dwellings as optimal material gathering grounds, or worse, that some human tissues would be naturally useful.
Further specialisations would occur in Guardians to specifically attack humans who are active threats. Units from different colonies would pass on information and designs to other colonies, if they exist which is likely at this point as some units would have fallen out of the communication net, when they randomly pass each other buy.
True A.I. would be highly unlikely to come about in this instance as self adaptation routines are unlikely to create anything more advanced than methods for dealing with issues. The necessity for actual thought and reason would simply not exist, and the ability to further improve on designs to achieve this end are equally unlikely. “Smart” behaviours would that mimick thought, at least in a tactical sense, would be present and slowly become more advanced. Electronics stores & warehouses will become hives. Places that humans have to go to, such as petrol station, will become assault grounds should humans become widely identified as a threat. In this instance, retaliation would occur, but be largely pointless as colonies and hives that are attacked will quickly learn to hide processing units and some basic HGs and Builders as a means of continuing the colony and starting over.
Whilst they may be rudimentary compared to the robotic swarms we expect from post-apocalyptic science fiction films, they still hold two major advantages over us. 1) They don’t work as collective individuals; and 2) they will have a learning time of zero for replacement units, versus the at least ten years it would take for our own. From a purely mathematical stand point, there’s no way they could lose.
